Limited Hadoop nodes as IBM Spectrum Scale nodes

For any deployment model, you do not have to put all the Hadoop cluster nodes as GPFS nodes.

  1. On all the Hadoop hosts that are either a NameNode or a DataNode in the native HDP cluster, assign a GPFS node so that the Transparency NameNodes and DataNodes are able to do a RPC in IBM Spectrum® Scale.
  2. If you need to configure additional Transparency DataNodes other than the native DataNodes, assign a GPFS Node on them as well.
  3. You could also have GPFS Nodes that do not use Transparency service. For example, if you want to use a host GPFS protocol node, assign GPFS Node on that host.